Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
PaddlePaddle
Serving
提交
503534a2
S
Serving
项目概览
PaddlePaddle
/
Serving
大约 1 年 前同步成功
通知
185
Star
833
Fork
253
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
105
列表
看板
标记
里程碑
合并请求
10
Wiki
2
Wiki
分析
仓库
DevOps
项目成员
Pages
S
Serving
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
105
Issue
105
列表
看板
标记
里程碑
合并请求
10
合并请求
10
Pages
分析
分析
仓库分析
DevOps
Wiki
2
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
提交
Issue看板
体验新版 GitCode,发现更多精彩内容 >>
提交
503534a2
编写于
7月 07, 2020
作者:
M
MRXLT
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
trt code
上级
2205dbed
变更
3
隐藏空白更改
内联
并排
Showing
3 changed file
with
70 addition
and
14 deletion
+70
-14
cmake/paddlepaddle.cmake
cmake/paddlepaddle.cmake
+12
-5
core/predictor/framework/infer.h
core/predictor/framework/infer.h
+5
-3
paddle_inference/inferencer-fluid-gpu/include/fluid_gpu_engine.h
...inference/inferencer-fluid-gpu/include/fluid_gpu_engine.h
+53
-6
未找到文件。
cmake/paddlepaddle.cmake
浏览文件 @
503534a2
...
...
@@ -31,10 +31,10 @@ message( "WITH_GPU = ${WITH_GPU}")
# Paddle Version should be one of:
# latest: latest develop build
# version number like 1.5.2
SET
(
PADDLE_VERSION
"1.
7.2
"
)
SET
(
PADDLE_VERSION
"1.
8.1
"
)
if
(
WITH_GPU
)
SET
(
PADDLE_LIB_VERSION
"
${
PADDLE_VERSION
}
-gpu-cuda
${
CUDA_VERSION_MAJOR
}
-cudnn7-avx-mkl"
)
SET
(
PADDLE_LIB_VERSION
"
${
PADDLE_VERSION
}
-gpu-cuda
10
-cudnn7-avx-mkl"
)
else
()
if
(
WITH_AVX
)
if
(
WITH_MKLML
)
...
...
@@ -92,8 +92,15 @@ LINK_DIRECTORIES(${PADDLE_INSTALL_DIR}/third_party/install/mkldnn/lib)
ADD_LIBRARY
(
openblas STATIC IMPORTED GLOBAL
)
SET_PROPERTY
(
TARGET openblas PROPERTY IMPORTED_LOCATION
${
PADDLE_INSTALL_DIR
}
/third_party/install/openblas/lib/libopenblas.a
)
ADD_LIBRARY
(
paddle_fluid STATIC IMPORTED GLOBAL
)
SET_PROPERTY
(
TARGET paddle_fluid PROPERTY IMPORTED_LOCATION
${
PADDLE_INSTALL_DIR
}
/lib/libpaddle_fluid.a
)
ADD_LIBRARY
(
paddle_fluid SHARED IMPORTED GLOBAL
)
SET_PROPERTY
(
TARGET paddle_fluid PROPERTY IMPORTED_LOCATION
${
PADDLE_INSTALL_DIR
}
/lib/libpaddle_fluid.so
)
ADD_LIBRARY
(
nvinfer SHARED IMPORTED GLOBAL
)
SET_PROPERTY
(
TARGET nvinfer PROPERTY IMPORTED_LOCATION /paddle/third_party/TensorRT-6.0.1.5/lib/libnvinfer.so
)
ADD_LIBRARY
(
nvinfer_plugin SHARED IMPORTED GLOBAL
)
SET_PROPERTY
(
TARGET nvinfer_plugin PROPERTY IMPORTED_LOCATION /paddle/third_party/TensorRT-6.0.1.5/lib/libnvinfer_plugin.so
)
ADD_LIBRARY
(
xxhash STATIC IMPORTED GLOBAL
)
SET_PROPERTY
(
TARGET xxhash PROPERTY IMPORTED_LOCATION
${
PADDLE_INSTALL_DIR
}
/third_party/install/xxhash/lib/libxxhash.a
)
...
...
@@ -101,4 +108,4 @@ SET_PROPERTY(TARGET xxhash PROPERTY IMPORTED_LOCATION ${PADDLE_INSTALL_DIR}/thir
LIST
(
APPEND external_project_dependencies paddle
)
LIST
(
APPEND paddle_depend_libs
xxhash
)
xxhash
nvinfer nvinfer_plugin
)
core/predictor/framework/infer.h
浏览文件 @
503534a2
...
...
@@ -553,7 +553,9 @@ class CloneDBReloadableInferEngine
};
template
<
typename
FluidFamilyCore
>
class
FluidInferEngine
:
public
CloneDBReloadableInferEngine
<
FluidFamilyCore
>
{
// class FluidInferEngine : public CloneDBReloadableInferEngine<FluidFamilyCore>
// {
class
FluidInferEngine
:
public
DBReloadableInferEngine
<
FluidFamilyCore
>
{
public:
FluidInferEngine
()
{}
~
FluidInferEngine
()
{}
...
...
@@ -603,13 +605,13 @@ class VersionedInferEngine : public InferEngine {
LOG
(
ERROR
)
<<
"Failed generate engine with type:"
<<
engine_type
;
return
-
1
;
}
VLOG
(
2
)
<<
"FL
A
GS_logtostderr "
<<
FLAGS_logtostderr
;
VLOG
(
2
)
<<
"FLGS_logtostderr "
<<
FLAGS_logtostderr
;
int
tmp
=
FLAGS_logtostderr
;
if
(
engine
->
proc_initialize
(
conf
,
version
)
!=
0
)
{
LOG
(
ERROR
)
<<
"Failed initialize engine, type:"
<<
engine_type
;
return
-
1
;
}
VLOG
(
2
)
<<
"FL
A
GS_logtostderr "
<<
FLAGS_logtostderr
;
VLOG
(
2
)
<<
"FLGS_logtostderr "
<<
FLAGS_logtostderr
;
FLAGS_logtostderr
=
tmp
;
auto
r
=
_versions
.
insert
(
std
::
make_pair
(
engine
->
version
(),
engine
));
if
(
!
r
.
second
)
{
...
...
paddle_inference/inferencer-fluid-gpu/include/fluid_gpu_engine.h
浏览文件 @
503534a2
...
...
@@ -190,7 +190,7 @@ class FluidGpuAnalysisDirCore : public FluidFamilyCore {
paddle
::
AnalysisConfig
analysis_config
;
analysis_config
.
SetModel
(
data_path
);
analysis_config
.
EnableUseGpu
(
100
,
FLAGS_gpuid
);
analysis_config
.
EnableUseGpu
(
1
5
00
,
FLAGS_gpuid
);
analysis_config
.
SwitchSpecifyInputNames
(
true
);
analysis_config
.
SetCpuMathLibraryNumThreads
(
1
);
...
...
@@ -198,11 +198,58 @@ class FluidGpuAnalysisDirCore : public FluidFamilyCore {
analysis_config
.
EnableMemoryOptim
();
}
if
(
params
.
enable_ir_optimization
())
{
analysis_config
.
SwitchIrOptim
(
true
);
}
else
{
analysis_config
.
SwitchIrOptim
(
false
);
}
/*
if (params.enable_ir_optimization()) {
analysis_config.SwitchIrOptim(true);
} else {
analysis_config.SwitchIrOptim(false);
}
*/
int
min_seq_len
=
1
;
int
max_seq_len
=
512
;
int
opt_seq_len
=
128
;
int
head_number
=
12
;
int
batch
=
50
;
std
::
vector
<
int
>
min_in_shape
=
{
batch
,
min_seq_len
,
1
};
std
::
vector
<
int
>
max_in_shape
=
{
batch
,
max_seq_len
,
1
};
std
::
vector
<
int
>
opt_in_shape
=
{
batch
,
opt_seq_len
,
1
};
std
::
string
input1_name
=
"src_text_a_ids"
;
std
::
string
input2_name
=
"pos_text_a_ids"
;
std
::
string
input3_name
=
"sent_text_a_ids"
;
std
::
string
input4_name
=
"stack_0.tmp_0"
;
std
::
map
<
std
::
string
,
std
::
vector
<
int
>>
min_input_shape
=
{
{
input1_name
,
min_in_shape
},
{
input2_name
,
min_in_shape
},
{
input3_name
,
min_in_shape
},
{
input4_name
,
{
batch
,
head_number
,
min_seq_len
,
min_seq_len
}},
};
std
::
map
<
std
::
string
,
std
::
vector
<
int
>>
max_input_shape
=
{
{
input1_name
,
max_in_shape
},
{
input2_name
,
max_in_shape
},
{
input3_name
,
max_in_shape
},
{
input4_name
,
{
batch
,
head_number
,
max_seq_len
,
max_seq_len
}},
};
std
::
map
<
std
::
string
,
std
::
vector
<
int
>>
opt_input_shape
=
{
{
input1_name
,
opt_in_shape
},
{
input2_name
,
opt_in_shape
},
{
input3_name
,
opt_in_shape
},
{
input4_name
,
{
batch
,
head_number
,
opt_seq_len
,
opt_seq_len
}},
};
analysis_config
.
EnableTensorRtEngine
(
1
<<
30
,
batch
,
5
,
paddle
::
AnalysisConfig
::
Precision
::
kHalf
,
true
,
true
);
analysis_config
.
SetTRTDynamicShapeInfo
(
min_input_shape
,
max_input_shape
,
opt_input_shape
);
AutoLock
lock
(
GlobalPaddleCreateMutex
::
instance
());
_core
=
...
...
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录